Uncertainty-Driven Adaptive Orchestration in Multi-Agent Retrieval-Augmented QA Systems

This project will build a new kind of AI question-answering system where several specialized “agents” work together under an adaptive controller. Instead of a single model answering a question in one shot, Omar’s work designs a coordinator that plans the steps, decides when more information is needed, and measures how confident the system should be before giving an answer. By using uncertainty signals and feedback from evidence, the controller can trigger extra retrieval or reasoning only when necessary, making answers both more reliable and more efficient. It also provides the student hands-on experience with advanced AI methods, supports thesis work, and strengthens long-term partnerships and joint publications between both institutions.

Faculty Supervisor:

Yassine Yaakoubi

Student:

Partner:

Tunisia Polytechnic School, University of Carthage

Discipline:

Computer science

Sector:

Artificial Intelligence; Information and Communications Technology (ICT); Technology

University:

Concordia University

Program: